Advancing Maritime Surveillance with UEIKAP: An Integrated AI-Based Framework for Wake Detection in Earth Observation Imagery

Detecting 'dark' ships, vessels operating without active Automatic Identification System (AIS) transponders or spoofing their signal to evade surveillance, is a critical challenge for maritime surveillance. The UEIKAP project (Unveil and Explore the In-depth Knowledge of Earth Observation data for maritime Applications) provides a new ship detection framework that aims to enhance maritime monitoring by focusing on the detection of wakes, instead of the ship itself. Unlike hull detection, which suffers from issues related to image resolution and is only useful to obtain the instantaneous position of vessels, wake detection provides information about vessel movement, offering insights about its velocity and heading. This project leverages a deep learning-based framework that integrates multiple data sources, including Earth Observation (EO) imagery, meteo-marine measurements, and AIS data. The fusion of such data is used to improve maritime monitoring and ship wake detection. The proposed system architecture comprises multiple modules dedicated to wake detection, including a detection model based on the YOLO family of object detection models and a classification model designed to differentiate between different sea surface phenomena with visible signatures in Synthetic Aperture Radar (SAR) imagery. The functionality of each module is demonstrated through representative examples. Initial results from the UEIKAP system demonstrate its effectiveness in detecting ship wakes under diverse environmental conditions. Performance metrics are analyzed, and the advantages of incorporating meteo-marine data are highlighted. The data fusion approach is shown to enhance maritime domain awareness by improving extraction of the navigational status of vessels and the characterization of the complex sea surface background.